feat(kap-server): accept secondary_model in the config API

[7Sageer]

Related Issue

No linked issue — the problem is explained below.

Problem

The engine supports the documented [secondary_model] configuration section, which lets newly spawned subagents bind to a separately configured model instead of inheriting the main agent's model. However, the server's configuration API (POST /api/v1/config) validates request bodies against a strict whitelist that does not include secondary_model, so API clients (web UI, extensions, external integrations) silently drop the field and the only way to set it is editing config.toml by hand.

What changed

• POST /api/v1/config now accepts secondary_model. The route's existing generic per-domain dispatch converts it to the secondaryModel config domain and routes it through the standard section validation and persistence, so [secondary_model] lands in config.toml and event.config.changed fires as for any other domain.
• GET /api/v1/config now hides the internal synthesized __secondary__ derived entry from the models view (it appears in the effective config whenever the recipe carries patch fields), matching how GET /models already filters it out of pickers. It can never persist — the config overlay strips it from models writes.
• Added route tests covering round-trip persistence of secondary_model and the __secondary__ filtering.

secondary_model request shape

For API callers: the field is a model pointer plus a patch.

• model — id of an entry in the configured models; newly spawned subagents bind to it.
• Every other field is a model override applied only to subagents, accepting the same fields as [models.*.overrides]: default_effort, off_effort, support_efforts, max_context_size, max_input_size, max_output_size, capabilities, display_name, reasoning_key, adaptive_thinking.
• Identity/connection fields (provider, api_key, base_url, …) do not belong here — they are inherited from the pointed entry. With no patch fields, subagents bind the pointed entry directly.

{ "secondary_model": { "model": "k3", "default_effort": "high", "max_output_size": 8192 } }

Responses echo the resolved recipe in camelCase (e.g. defaultEffort). Field reference: config docs.

Known limitation (pre-existing, unchanged): the session secondary-model warning is computed once when the main agent is created, so updating the section mid-session does not refresh GET /sessions/{id}/warnings; spawn-time resolution remains the backstop.

Checklist

• I have read the CONTRIBUTING document.
• I have linked a related issue, or explained the problem above.
• I have added tests that prove my feature works.
• Ran gen-changesets skill, or this PR needs no changeset.
• Ran gen-docs skill, or this PR needs no doc update.

[changeset-bot]

🦋 Changeset detected

Latest commit: cacecbe

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[pkg-pr-new]

pnpm dlx https://pkg.pr.new/@moonshot-ai/kimi-code@cacecbe

npx https://pkg.pr.new/@moonshot-ai/kimi-code@cacecbe

commit: cacecbe